Active Site Reactant Center Geometry in the CoII−Product Radical Pair

Active Site Reactant Center Geometry in the CoII−Product Radical Pair State of ... Simulation of the EPR line shape yielded a best-fit CoII−C2 dis...
0 downloads 0 Views 406KB Size
Download PDF
J. Phys. Chem. B 2005, 109, 3053-3064

3053

Active Site Reactant Center Geometry in the CoII-Product Radical Pair State of Coenzyme B12-Dependent Ethanolamine Deaminase Determined by Using Orientation-Selection Electron Spin-Echo Envelope Modulation Spectroscopy Jeffrey M. Canfield and Kurt Warncke* Department of Physics, Emory UniVersity, Atlanta, Georgia 30322 ReceiVed: August 24, 2004; In Final Form: NoVember 10, 2004

The distances and orientations among reactant centers in the active site of coenzyme B12-dependent ethanolamine deaminase from Salmonella typhimurium have been characterized in the CoII-product radical pair state by using X-band electron paramagnetic resonance (EPR) and two-pulse electron spin-echo envelope modulation (ESEEM) spectroscopies in the disordered solid state. The unpaired electron spin in the product radical is localized on C2. Our approach is based on the orientation-selection created in the EPR spectrum of the biradical by the axial electron-electron dipolar interaction. Simulation of the EPR line shape yielded a best-fit CoII-C2 distance of 9.3 Å. ESEEM spectroscopy performed at four magnetic field values addressed the hyperfine coupling of the unpaired electron spin on C2 with 2H in the C5′ methyl group of 5′-deoxyadenosine and in the β-2H position at C1 of the radical. Global ESEEM simulations (over the four magnetic fields) were weighted by the orientation dependence of the EPR line shape. A Nelder-Mead direct search fitting algorithm was used to optimize the simulations. The results lead to a partial model of the active site, in which C5′ is located a perpendicular distance of 1.6 Å from the CoII-C2 axis, at distances of 6.3 and 3.5 Å from CoII and C2, respectively. The van der Waals contact of the C5′-methyl group and C2 indicates that C5′ remains close to the radical species during the rearrangement step. The C2-Hs-C5′ angle including the strongly coupled hydrogen, Hs, and the C5′-Hs orientation relative to the C1-C2 axis are consistent with a linear hydrogen atom transfer coordinate and an in-line acceptor p-orbital orientation. The trigonal plane of the C2 atom defines sub-spaces within the active site for C5′ radical migration and hydrogen atom transfers (side of the plane facing CoII) and amine migration (side of the plane facing away from CoII).

Introduction Ethanolamine deaminase is a bacterial coenzyme B12 (adenosylcobalamin, Figure 1) dependent enzyme1 (also known as ethanolamine-ammonia lyase) that catalyzes the elimination of ammonia from aminoethanol 1 to produce acetaldehyde 4.2 The key steps in the reaction are mediated by radical species. As depicted in Figure 2, homolytic cleavage of the cobaltcarbon bond of coenzyme B12 produces low-spin (S ) 1/2) CoII in cobalamin, and a C5′-centered 5′-deoxyadenosyl radical species that is proposed to initiate catalysis in all coenzyme B12-dependent enzymes.3-5 Hydrogen atom abstraction from the carbinol carbon (C1) of the bound substrate by the C5′ radical generates the C1-centered substrate radical 2,6,7 and the C5′methyl group in 5′-deoxyadenosine. The substrate radical then rearranges to a product radical, with the unpaired electron spin density localized on C2.8 Subsequent hydrogen atom transfer from the C5′ methyl group to C2 of the product radical yields a diamagnetic product species and re-forms the 5′-deoxyadenosyl radical. The 5′-deoxyadenosyl radical and CoII may then recombine.9 The structures of five coenzyme B12-dependent enzymes, including free and substrate-, inhibitor-, or productbound states, have been determined by X-ray crystallography.10-19 These states may contain CoII,20 but apparently not organic radical species. Structural information about the reactant centers in the unstable CoII-radical pair intermediate states has been * Corresponding author. Tel: 404-727-2975. Fax: 404-727-0873. E-mail: [email protected].

revealed by high-resolution techniques of electron paramagnetic resonance (EPR) spectroscopy, applied to disordered samples of cryotrapped CoII-radical pair intermediates.21-23 Here, we apply the combined experimental and simulation approach of orientation-selection electron spin-echo envelope modulation (ESEEM)23 spectroscopy to determine the three-dimensional geometry of CoII, the C5′-methyl group and the C1-C2 product radical frame in the CoII-product radical pair intermediate in ethanolamine deaminase from Salmonella typhimurium.24,25 The product radical accumulates during steady-state turnover on the native substrate, aminoethanol, and is stabilized for lowtemperature spectroscopy by cryotrapping.8 EPR simulation analysis of the product radical line shape showed that the CoIIC2 distance in the CoII-product radical pair is 9.7 Å,26 which is comparable to the best-fit value of 9.3 Å (range 8.5-10.5 Å) that we report here. The CoII-C2 distance is slightly shorter than the CoII-C1 distances of 10-12 Å,27 11 Å,7 and 11.1 Å (range 10.9-11.5 Å)23 reported by different groups for the CoIIsubstrate radical pair state. A carbinolamine (1-aminoethan-1ol-2-yl) structure 3 for the product radical was determined from differences in the EPR line shape of the product radical generated from 14N- or 15N-labeled aminoethanol.28 The radical rearrangement step therefore proceeds by the amine migration pathway, in which the amine nitrogen migrates from C2 to C1 to form 3.28 Multifrequency three-pulse powder ESEEM of the hyperfine interactions between the product radical and the 2Hlabeled C5′-methyl group showed coupling with the C5′-2H at effective separations of 2.3 Å (one 2H) and 4.2 Å (two 2H).29

10.1021/jp046167m CCC: $30.25 © 2005 American Chemical Society Published on Web 01/25/2005

3054 J. Phys. Chem. B, Vol. 109, No. 7, 2005

Figure 1. Depiction of the structure of vitamin B12 coenzyme [adenosylcob(III)alamin or adenosylcobalamin]. The X-ray crystallographic structures of cobalamins have been reviewed.66,67 The β-axial ligand is the 5′-deoxyadenosyl group (above page plane) and the R-axial ligand is 5,6-dimethylbenzimidazole (below page plane). The coenzyme retains dimethylbenzimidazole as an R-axial ligand when bound in ethanolamine deaminase.68,69 R1 and R2 refer to acetamide and propionamide side chains. The C5′ carbon of 5′-deoxyadenosyl is labeled, and the two exchangeable hydrogen atoms on the coenzyme are shown in bold.

CHART 1

These results led to a model in which the C2-C5′ distance was 3.3 ( 0.1 Å.29 Hyperfine interactions of the C2 unpaired electron spin with the β-2H at C1 revealed two types of β-2H coupling, which led to the proposal of two different C1-C2 rotamer states for the product radical.28,29 The powder three-pulse ESEEM did not, however, report on the three-dimensional orientation among CoII, C2, C1, and the C5′ methyl group, which remains to be elucidated. We have introduced and described the approach of orientation-selection spectroscopy for the case of radical pair systems,23 in which the orientation-selection in the EPR line shape arises from the electron-electron dipolar interaction. Different magnetic field values across the line shape correspond to subpopulations of orientations of the electron-electron vector (R) relative to the external magnetic field vector (B0). The orientations contributing to the spectrum at any magnetic field value can be calculated from simulations of the EPR spectrum. ESEEM collected at a particular magnetic field is representative of the corresponding subpopulation of orientations. The electronnuclear interactions that give rise to the ESEEM include an anisotropic contribution from the dipolar interaction that is

Canfield and Warncke

Figure 2. Minimal mechanism for catalysis in vitamin B12 coenzymedependent enzymes.5 The forward direction of reaction is indicated by the arrows. In ethanolamine deaminase, the detectable reversibility of individual steps varies among different substrates.70 The brackets represent the active site region in the protein interior. Substrate-derived species are designated S-H (substrate), S• (substrate radical), P• (product radical), PH′, PH′′ (bound product species), and PH (free products). The 5′-deoxyadenosyl β-axial ligand is represented as AdCH2- in the intact coenzyme, and as Ad-CH2• (5′-deoxyadenosyl radical) or Ad-CH3 (5′-deoxyadenosine) following cobalt-carbon bond cleavage. The cobalt ion and its formal oxidation states are depicted, but the corrin ring and R-axial ligand of the coenzyme are not shown for clarity. Steps of the cycle are denoted RPS (radical pair separation), HT1 (first hydrogen atom transfer), RR (radical rearrangement), HT2 (second hydrogen atom transfer), RPR (radical pair recombination), and SB/PR (substrate binding/product release).

sensitive to the orientation of the electron-nuclear vector relative to the external magnetic field. Therefore, the orientation of the electron-electron and electron-nuclear vectors can be determined by using the EPR simulations to weight the contribution of different orientations to the ESEEM. Global simulation of the ESEEM collected at different magnetic field values then reveals the orientations of the nuclei relative to the electron-electron axis. We have implemented a Nelder-Mead Simplex direct search optimization algorithm30,31 to automate fitting. The experimental and theoretical approaches are related to the orientation-selection ENDOR32-35 and ESEEM36 spectroscopies in transition metal systems, where the orientation selection is created primarily by anisotropy in the g-tensor. For the present studies, the product radical was cryotrapped during steady-state turnover on aminoethanol-1,1,2,2-2H4 (2H4aminoethanol) or natural abundance (1H4-)aminoethanol. The multiple turnovers of the enzyme prior to sample cryotrapping result in the rapid and complete exchange of 2H into all three sites of the C5′ methyl group of 5′-deoxyadenosine.37 In addition, the nonsolvent-exchangeable hydrogen atoms on the product radical (1,1,2,2-positions) are 2H labeled. The hyperfine coupling parameters for the three C5′-2H and the β-2H have been previously determined in the powder three-pulse ESEEM study.29 The strong coupling of the two equivalent R-deuterons38 places it outside of the microwave pulse bandwidth detection limit of the ESEEM technique.39-41 In addition, the three-pulse study established that the stronger of the two β-2H hyperfine

Product Radical Structure in Ethanolamine Deaminase

J. Phys. Chem. B, Vol. 109, No. 7, 2005 3055

couplings is detected at a microwave frequency/magnetic field value of 10.89 GHz/388.0 mT but is not detected above the noise level at the lower X-band frequency/field of 8.86 GHz/ 313.0 mT.29 In concurrence with this result, the stronger β-2H coupling (denoted Hβ2) was not observed in the two-pulse ESEEM studies performed at 8.77 GHz that are reported here, whereas features from the weaker β-2H coupling (denoted Hβ1) are prominent. The magnetic interactions of the unpaired electron at C2 with 2H nuclei were measured at four magnetic field values by using the two-pulse ESEEM technique.39-41 Global simulation analysis of the 2H-ESEEM reveals the orientation of the three C5′ methyl hydrogen and β-hydrogen positions relative to the CoII-C2 axis. A model for the three-dimensional geometry of the 2H nuclei, C2, C1, C5′, and CoII is presented. The model provides insights into the mechanism of radical-mediated catalysis in ethanolamine deaminase. Experimental Procedures Enzyme Preparation. Enzyme was purified from the Eschericia coli overexpression strain incorporating the cloned S. typhimurium ethanolamine deaminase coding sequences24 essentially as described,24 with the exception that the enzyme was dialyzed against buffer containing 100 mM HEPES (pH 7.45), 10 mM KCl, 5 mM dithiothreitol, 10 mM urea, and 10% glycerol.42 Enzyme activity was determined as described43 by using the coupled assay with alcohol dehydrogenase/NADH. The specific activity of the purified enzyme with aminoethanol as substrate was 25-35 µmol/min/mg. Sample Preparation. Adenosylcobalamin (Sigma Chemical Co.), aminoethanol-1,1,2,2-2H4 (2H4-aminoethanol; Cambridge Isotope Laboratories, Inc.) and natural abundance aminoethanol (Aldrich Chemical Co.) were purchased from commercial sources. The reactions were performed in air-saturated buffer containing 100 mM HEPES (pH 7.5), 10 mM KCl, and 5 mM dithiothreitol. Identical results were obtained with air-saturated and anaerobic samples. All manipulations were carried out on ice under red safe-lighting. The product radical was generated by using a procedure for fast cryotrapping of steady-state intermediate states in ethanolamine deaminase.8 The final concentration of enzyme was 10-15 mg/mL, which is equivalent to 20-30 µM for a holoenzyme molecular mass of 500 000 g/mol.25 Adenosylcobalamin was added to 120-180 µM, which is stoichiometric with active sites. The active site/holoenzyme stoichiometry of 6 is based on adenosylcobalamin titrations of substrate radical formation (K. Warncke, unpublished) and is in agreement with the value obtained by two separate methods.44,45 The procedure for specific incorporation of 2H into the hydrogen positions of the C5′ carbon in enzyme-bound adenosylcobalamin has been described in detail.22 Turnover of the enzyme on 2H4-aminoethanol leads to exchange of 2H into the C5′ hydrogen positions of the coenzyme.22 Following mixing of the enzyme-substrate solution with adenosylcobalamin, the sample was loaded into a 4 mm o.d. EPR tube, and the tube was plunged into liquid nitrogen-chilled isopentane (T ≈ 130 K) to trap the CoII-product radical pair state. The total elapsed time from mixing to isopentane immersion was approximately 15 s. Continuous-Wave EPR Spectroscopy. EPR spectra were obtained by using a Bruker ER200D EPR spectrometer equipped with a Bruker 4102ST TE102 cavity, HP 4256L frequency counter, Varian V3603 electromagnet, and Fieldial Mark I regulator/power supply, and Air Products cryostat and temperature controller modified for nitrogen gas flow sample cooling.

Figure 3. Coordinate systems used to define the relative positions of CoII, C2, C1, C5′, and the C5′ methyl hydrogen atoms, and the R- and β-hydrogen atoms of the product radical in the active site of ethanolamine deaminase. (A) The xcyczc coordinate system used to define the orientation of the magnetic field vector (B0) with polar and azimuthal angles, θ and φ, relative to the active site in EPR and ESEEM simulations. Hydrogen site, Hn, lies a distance, rn, from the unpaired electron spin on C2, and βn is the angle between the C2-Hn bond axis and the C2-CoII axis. (B) The x2y2z2 coordinate system, which positions the C2-C1 bond axis and the C2 p-orbital relative to C5′ and CoII. (C) Detail of the x2y2z2 coordinate system, showing the positions of the R- and β-hydrogens on the product radical. (D) The x5y5z5 coordinate system, which positions the C5′-methyl hydrogen sites relative to the C5′ center. The open circle shows the position of a fourth hydrogen position in an open tetrahedron, which is occupied by the ribose C4′ atom in the 5′-deoxyadenosyl group.

ESEEM Data Acquisition and Processing. ESEEM was collected by using the two-pulse microwave pulse sequence.39-41 Envelope modulation was dead time-reconstructed46 and cosine Fourier transformed to generate ESEEM frequency spectra. All data processing was done using locally written Matlab code (Matlab v. 6.5 and Matlab Compiler v. 3.0) on Windows PC’s. Theory Coordinate Systems. Figure 3 shows the coordinate systems that were used to specify the orientations and separation distances among the atom centers. The xcyczc coordinate system used for EPR and ESEEM simulations is shown in Figure 3A. The polar (θ) and azimuthal (φ) angles describe the orientation of the magnetic field vector B0 in spherical coordinates. The coupled hydrogen sites (either 2H or 1H) are denoted as Hn in the calculations, where n ) 1-7. The three C5′ methyl hydrogen sites correspond to n ) 1-3 and are denoted in Figure 3 as Hm1, Hm2, and Hm3. The two R-H bonded to C2 are denoted as HR1 and HR2, and the two types of β-H bonded to C1 are denoted as Hβ1 and Hβ2. Each Hn is treated as if it lies in the xczc plane a distance rn from C2, because the EPR spin Hamiltonian (eq

3056 J. Phys. Chem. B, Vol. 109, No. 7, 2005

Canfield and Warncke

10) is axial about zc and all orientations θ, φ of B0 are sampled in the powder average. The Hn-C2-CoII angle for each hydrogen is denoted βn. The C2-CoII vector can be reversed without altering the simulated ESEEM, so that βn and 180° βn give the same ESEEM waveform. Therefore, only βn values from 0 to 90° are considered. To obtain the βn and rn values for ESEEM calculations, the atom positions are calculated and the following expressions are used:

8 rn ) |9 C2-Hn |

(1)

II Co B -C2 ‚C2-H B n

II B -C2 ‚C2-H B Co n (2) -cos(βn) ) ) II R‚r |Co B -C2 ||C2-H B n n|

Additional coordinate systems are used to position the different Hn sites on the product radical and C5′ methyl group. Figure 3B shows the x2y2z2 coordinate system, which positions C5′, CoII, C1, C2 and the lobes (denoted “+p” and “-p”) of the p-orbital on C2. Figure 3C presents another view of the x2y2z2 coordinate system, which highlights the positions of the product radical atoms relative to C2. The two R-hydrogens (HR1 and HR2) on C2 are placed in the x2y2 plane and oriented symmetrically about the y2z2 plane. Two different β-hydrogen sites28 are defined, as follows: Hβ1, with x2 g 0 and Hβ2, with x2 e 0, both with z2 e 0 and y2 g 1.44 Å. This constrains C2-Hβn to a range of 1.81-2.54 Å, because standard bond lengths are used for C1-C2 (1.44 Å) and C-H (1.1 Å) in the product radical. The positions of the Hβn in the x2y2z2 coordinate system were determined from the Hβn isotropic hyperfine coupling. This determination, and the origin of the two β-hydrogen sites, is described under ESEEM simulations, below. Figure 3D shows the orientation of the C5′ methyl group H atoms within the x5y5z5 coordinate system. The C5′-Hmn distances and inter-hydrogen bond angles are fixed at 1.1 Å and 109.5°, respectively, corresponding to sp3-hybridization at C5′. In Figure 3D, the fourth C5′ bonding position is shown to be occupied by the C4′ carbon of the ribose moiety. Powder Averaging and Angle Selection. Powder averaging was used in simulations of the EPR and ESEEM spectra from the frozen solutions used in this study. If f(θ,φ) is the EPR spectrum that a single crystal would give for B0 oriented in the θ, φ direction with respect to the crystal’s xcyczc axes, then the powder average is given by

∫0πsin(θ) dθ∫02πdφ f(θ,φ) 〈f(θ,φ)〉 ) ∫0πsin(θ) dθ∫02πdφ

(4)

To account for the orientation selection of the ESEEM, the factor, P(θ,φ), which reflects the relative population or relative EPR absorption transition intensity, at each orientation, θ, φ, enters eq 4 as follows:

(5)

In practice, the orientation averaging is approximated numerically by summing over a finite set of orientations θj, φj.23 Fitting Procedures. The experimental EPR and ESEEM spectra were fitted by using the Nelder-Mead simplex direct search method30,31 as implemented in Matlab’s “fminsearch” algorithm. This algorithm minimizes a fitting function, σ, that reflects the goodness-of-fit between an experimental curve (the data points Yj) and a calculated curve (the points xj). The calculated curve is given by the following general expression

yj ) Axj + B

(6)

In eq 6, the coefficient A scales the range of xj whereas B accounts for a constant “baseline” contribution. The fitting function, σ, is given by the squared deviation of Yj and yj, as follows:

σ)

∑j (Yj - yj)2 ) ∑j [Yj - (Axj + B)]2

(7)

The following first derivative expressions for dσ/dA and dσ/dB were solved to find the set of A, B values that gave the smallest σ value for a particular set of experimental Yj and calculated xj.

dσ

)

dA

∑j (2Axj2 - 2xjYj + 2Bxj) ) 0

where

d 2σ dA dσ dB

)

∑j (2B - 2Yj + 2Axj) ) 0

)

2

∑j (2xj2) g 0

(8A)

where d2σ dB2

)

∑j (2) g 0

(8B)

Details of the fitting procedure have been presented previously.23 To fit the experimental 2H/1H quotient ESEEM waveforms, a modification of eqs 6-8 was used. The calculated curve, zj(τj), where τj is the dephasing time in the two-pulsed ESE sequence, was multiplied by ekτj to obtain

(3)

Similarly, if F(θ,φ) is the ESEEM waveform that a single crystal would give, the powder-averaged ESEEM waveform is given by an expression comparable to eq 3, as follows:

∫0πsin(θ) dθ∫02πdφ F(θ,φ) 〈F(θ,φ)〉 ) ∫0πsin(θ) dθ∫02πdφ

∫0πsin(θ) dθ∫02πdφ F(θ,φ) P(θ,φ) 〈F(θ,φ)〉 ) ∫0πsin(θ) dθ∫02πdφ P(θ,φ)

xj ) zj(τj)ekτj

(9)

The factor ekτj corrects for any differences between the individual and 1H echo envelopes caused by irreversible loss of spin coherence. In the individual envelopes, the envelope decay is quasi-exponential and governed by the phase memory time, TM. Thus, for a 2H/1H quotient ESEEM waveform, k ) (1/TM,1H)(1/TM,2H). EPR Simulations. The program MENO47 was used to simulate the EPR spectra. The program treats spin Hamiltonians, such as those previously reported for radical pairs in different coenzyme B12-dependent enzymes,27 that contain axial dipolar interaction terms, isotropic exchange terms (J), an axial g tensor, and axial hyperfine tensors (A). The Hamiltonian in the xcyczc coordinate system is written as follows, 2H

Product Radical Structure in Ethanolamine Deaminase

J. Phys. Chem. B, Vol. 109, No. 7, 2005 3057

H ) h[-JS B1‚S B2 + A⊥(S1xI1x + S1yI1y) + A|S1zI1z] + µ0 b b1‚R µ 1‚µ b2 3(µ B)(µ b2‚R B) µ 2)‚B B0 + (10) (µ b1 + b 3 5 4π R R

[

]

where

b µ j ) βe(gjxSjxxˆ + gjySjyyˆ + gjzSjzzˆ)

(11)

The subscripts “1” and “2” correspond to CoII and the radical, respectively. In the simulations, R is along zc, g1x ) g1y ) g⊥ ) 2.26 and g1z ) g| ) 2.01, g2xyz ) 2,27,45 S1 ) S2 ) 1/2, and I1 ) 7/2. The cobalt hyperfine parameters are A⊥ ) 30 MHz (10 × 10-4 cm-1) and A| ) 309 MHz (103 × 10-4 cm-1).27 The sign convention for J in MENO is that J < 0 indicates antiferromagnetic coupling and thus corresponds to a lower energy for the singlet relative to the triplet state.47 Different isotropic Gaussian line widths were used for the CoII and radical contributions to the spectrum to account for the different relaxation behavior and unresolved hyperfine couplings for each electron.6 To optimize the fit, J, R, and the line widths were varied from a variety of initial parameter sets to minimize σ as specified in eq 7, with Yj from the experimental spectra and xj from the direct output from MENO. Although both J and R contribute to the magnitude of the doublet splitting in the radical line shape, R also determines the anisotropic part of the dipolar interaction, and thus influences the line width of each radical doublet feature. Agreement between simulations performed with MENO, and simulations performed with the frequency-based, direct-diagonalization method of the program, DSNANI,48 shows that the perturbation series used to derive MENO holds well for the radical pair Hamiltonian used here and in previous studies.23 A series of EPR absorption spectra were calculated for each θ value, and cross-sections of the EPR transition intensity versus orientation at the particular values of 306.5, 310.5, 313.3, and 319.0 mT were taken. The corresponding spectra and crosssections are shown in Figures 5 and 6, respectively. The cross section at each field was rescaled to have a maximum of 1, and each resulting cross section (as shown in Figure 6) was denoted as P(θ,φ). Because the area under an EPR absorption spectrum can be used to measure the spin population,38 the resultant P(θ,φ) curves were then used to weight the powder average, as given in eq 5, when calculating the ESEEM spectra. In the absence of orientation selection, P(θ,φ) ) 1 would hold, and a normal powder average, as given by eq 4, would result. ESEEM Simulations. IndiVidual Electron-Nuclear Couplings. The ESEEM simulation approach was based on Mims’ formalism.49,50 The implementation of this approach has been described.22,51 The simulations focused on the coupling of the unpaired electron spin on C2 with hydrogen sites on the C5′ methyl group and on the product radical. Each hydrogen site, Hn, was parametrized by rn and βn, as defined in Figure 3A, and Aiso,n, the isotropic hyperfine coupling. The ESEEM from hydrogen site Hn was calculated by using the spin Hamiltonian w

B0‚S B2 - βNgNB B0‚BI 2 + hS B2‚A2‚BI 2 H ) βegeB

(12)

where ge ) 2, S2 ) 1/2, gN ) 5.5857 for 1H and 0.8574 for 2H, w I2 ) 1/2 for 1H and 1 for 2H, and A 2 is the complete hyperfine tensor. The nuclear quadrupole interaction terms for the I2 ) 1 2H nucleus were neglected because they are generally small compared to the hyperfine couplings and had little influence on simulations in preliminary work.

TABLE 1: Best-Fit EPR Simulation Parameters and Fit Numbers input parameters J/10-4

-1

cm J/MHz R/Å radical line width/mT CoII line width/mT g(radical) g⊥(CoII) g|(CoII) A⊥/10-4 cm-1 (CoII) A|/10-4 cm-1 (CoII)

best-fit valuesa -67.2 -202 9.26 5.16 6.08 2.000 2.260 2.010 10 103

fitting parameters

best-fit values

σEPR/1029

1.19

best-fit rangeb -70.8 -212 8.49 4.79 4.53 1.998 2.249 1.987 4.78 88.95

-63.7 -191 10.50 5.67 8.03 2.001 2.265 2.032 14.77 105.72

best-fit range 1.19

2.00

a

The best-fit values were obtained by optimizing J, R, and the line widths, with g value and hyperfine parameters held fixed at literature values. b The minimum and maximum values for each parameter represent the range obtained when the indicated parameter was varied with all of the other parameters held fixed at their best-fit values. These extreme values reflect the parameter range in which σEPR remained e2.00 × 1029.

The model of Gordy52 for the interaction between an unpaired electron spin in a p-orbital and a nuclear spin was adapted for the treatment of the hyperfine interactions, with the exception of the R-H. In the Gordy model, the electron spin density is concentrated equally at effective centers of the p-orbital lobes, 0.72 Å above and below the nodal plane, and the dipole interactions between Hn and each lobe are summed. The complete (including both isotropic and dipolar contributions) rhombic tensor for the R-hydrogen hyperfine couplings was given explicitly (A2xyz ) [-58, -91, -29] MHz for 1H coupling and A2xyz ) (gD/gH)[-58, -91, -29] MHz for 2H coupling, where gD ) 0.8574 and gH ) 5.5857.38 The Aiso,β2 and rβ2 parameters were also given explicitly by the values determined by multifrequency three-pulse ESEEM (see Table 2).28,29 The Hβ2 parameters were not adjustable, because the ESEEM from the Hβ2 coupling is not detectable above the noise level at the microwave frequency/magnetic field values used in the present study.29 The ESEEM from the Hβ1 coupling was observed, as expected,29 and therefore, the Aiso,β1 and rβ1 parameters were adjustable. For each site Hβn, the dihedral angle between the p-orbital axis and the C1-Hβn bond, θβn, was calculated from the fitted value of the isotropic hyperfine coupling by using the Heller-McConnell relation:53 Aiso,βn ) FB2 cos2 θβn, where F ) 1 is the unpaired electron spin density at C2,8 and B2 ) 24.9 MHz for 2H.54 w The hyperfine tensor for each Hn site, A 2, was rotated into the xcyczc coordinate system, and the ESEEM from each orientation was weighted as in eq 5 by using P(θ), the θ-dependence obtained from the EPR simulation. This allowed the treatment of effects on the ESEEM caused by J and R without explicitly including these terms in the spin Hamiltonian of eq 12. The powder average for each of the seven individual coupled hydrogen sites, Hn (in both the 2H and 1H occupied states), was calculated independently, as in eq 5.39,41,55 Owing to the large number of powder average orientations to sample for each Hn site and multiple parameters to vary during optimization, Mims’ expressions49 for the envelope modulation from a single site at a single orientation were rearranged into the following form:

E(τ) )

∑q Aqeiω τ q

(13)

where Aq includes τ-independent amplitude terms and eiωqτ is

3058 J. Phys. Chem. B, Vol. 109, No. 7, 2005

Canfield and Warncke

TABLE 2: Best-Fit ESEEM Simulation Parameters and Ranges Obtained from Global Fitting of the ESEEM Collected at Four Different Magnetic Field Valuesa variable

best-fit value

R/Å r(C2-C5′)/Å r(CoII-C5′)/Åe C5′-Ha-C2/deg Aiso,a/MHz Aiso,β1/MHz Aiso,β2/MHz ra/Åb rb/Åb rc/Åb rβ1/Åc rβ2/Åd βa/degb βb/degb βc/degb ββ1/degc ββ2/degd

9.26 3.48 6.34 12.47 161 -0.75 4.57 7.8 2.42 4.07 4.09 2.44 2.37 20.8 21.6 26.1 56.2 62.3

global fit number

best-fit value

(Σσt)(Πσf)/10

14

1.86

best-fit range fixed 3.05 5.88 12.18 143 -0.79 4.48

3.55 6.82 12.67 180 -0.66 4.66 fixed

2.09 3.67 3.72 2.37

2.46 4.11 4.21 2.52 fixed

2.51 0.103 0.199 44.4 38.6

29.5 42.8 39.7 83.3 89.9

ESEEM studies28,29 have shown that there are two populations of active sites that are distinguished by having different β-2H hyperfine couplings. There is one type of β-H hyperfine coupling in each active site. The coupling with the C5′ methyl group D appears to be identical in each type of site. The terms Eβ1 and D Eβ2 represent the modulation from the β-2H coupling in the two types of active site, and pβ1 and pβ2 are the corresponding relative populations, where pβ1 + pβ2 ) 1. The term ΠEX represents modulation from hyperfine coupling to nuclei other than the hydrogen sites defined above. The overall ESEEM waveform from the sample prepared by using natural abundance 1H -aminoethanol is given by the following expression: 4 H H EHtot ) EHm1EHm2EHm3EHR1EHR2[pβ1Eβ1 + pβ2Eβ2 ]

2.17

a

The best-fit range column lists ranges from more than 70 000 separate fitting attempts (simulation runs). From the best-fit global fit number corresponding to the results in Figures 7 and 8, a best fit range (listed at the bottom of the Best Fit Range column) was selected. Simulation runs with global fit numbers within this range were used to determine the ranges for each parameter. Thus, if r(C2-C5′) is in the range listed, a global fit with a fit number within the best-fit range can be achieved by using a combination of all of the other fitting parameters from within their respective ranges. Plots of the fit values versus the range for different parameters are shown in Figure S1. b The ra, rb, rc and βa, βb, βc parameters are related to the rm1, rm2, rm3 and βm1, βm2, βm3 parameters but sorted so that ra e rb e rc holds. Thus, Ha is the strongly coupled hydrogen, Hs, and Hb and Hc are the weakly coupled hydrogens, Hw. c The Aiso,β1, rβ1, and ββ1 parameters refer to the observed β-2H coupling with Hβ1, with pβ1 ) pβ2 ) 0.5, as given in eqs 14-16. d The Aiso,β2, rβ2, and ββ2 parameters refer to the population of β-2H coupling with Hβ2, which is not observed under the experimental conditions used here. The rβ2 parameter is fixed by the value reported in ref 29, and pβ1 ) pβ2 ) 0.5, as given in eqs 14-16. e Inversion symmetry of the model leads to two possible values for each of the CoII-C5′ and CoII-C1 distances.

the τ-dependent propagator.39,41,56 In this form, the exponential terms are evaluated as infrequently as possible over the range of τ values. The Matlab Compiler (v. 3.0) was also used to speed this portion of the code, which was the main bottleneck for the calculations. Combination of Electron-Nuclear Couplings. The total twopulse ESEEM for a group of nuclei interacting with the same electron is given by the product of the ESEEM obtained separately from each nucleus.49,57 Thus, the simulated total ESEEM waveform was obtained by powder-averaging the ESEEM (eq 5) from each nucleus separately, followed by multiplication of the individual powder-averaged ESEEM waveforms.58,59 The total ESEEM waveform from the 2H-labeled sample is given as follows: D D EDtot ) EDm1EDm2EDm3EDR1EDR2[pβ1Eβ1 + pβ2Eβ2 ]

Ex ∏ x

(14)

where D represents 2H occupancy of a hydrogen site. EDRn is the waveform from coupling to the nth R-2H site on C2, EDmn is the ESEEM waveform from coupling to the nth C5′ methyl D hydrogen site, and Eβn is the waveform from coupling to the nth β-2H site on C1. Previous multifrequency three-pulse

(15)

where H represents 1H occupancy of the hydrogen sites defined above. Envelope-dividing eq 14 by eq 15 gives the following total ESEEM quotient waveform:

best-fit range 1.86

Ex ∏ x

EQtot )

D D EDm1EDm2EDm3EDR1EDR2[pβ1Eβ1 + pβ2Eβ2 ] H H EHm1EHm2EHm3EHR1EHR2[pβ1Eβ1 + pβ2Eβ2 ]

(16)

Note that ΠEX is eliminated in eq 16, so that the quotient modulation does not include interfering contributions from hyperfine couplings that are common to the 2H- and 1H-labeled product radical states. ESEEM Fitting. For a particular pair of experimental and calculated waveforms at the same magnetic field, a set of A and k values could be determined that gave the minimum possible σ value for the waveform fit, denoted στ. In the determination of στ, a value of B ) 0 was used. The 250 τ values from 192 to 1686 ns were used to calculate στ, which included the dominant modulation amplitude and avoided the poor signal-to-noise ratio region at τ > 1686 ns. The waveform was further processed by removing a best-fit first-order polynomial from the data, followed by reconstruction of the dead time.46 ESEEM frequency spectra were obtained from the processed waveform by discrete Fourier transformation by using Matlab’s fft algorithm, and taking the real part (cosine Fourier transform) of the result.39,55 Processed experimental data for τ values from 0 to 2274 ns were used to calculate the frequency domain spectra. No rescaling or offsets of the Fourier transforms were performed. Therefore, in the calculation of the minimum σ for fitting the Fourier transforms, σf, values of A ) 1 and B ) 0 were used. The 117 frequencies from 0.49 to 9.9 MHz were used in calculating σf, because the Fourier transform was distorted by truncation artifacts near the origin and the range of frequencies 1.6 µs, and this large relative noise at large τ is enhanced by the envelope division. The differences among the ESEEM collected at different magnetic field values show that orientation selection is present. The cosine Fourier transforms of the ESEEM shown in Figure 7 are presented in Figure 8. The intensity in the spectral region e10 MHz is dominated by features arising from 2H hyperfine coupling. Significant intensity from the 1H coupling appears at ν1H (13-14 MHz) and 2ν1H (26-27 MHz) and is therefore not shown in the spectral range of Figure 8. The amplitudes of features in the ESEEM spectra vary significantly with magnetic field, but the frequency ranges of the transitions are comparable, and in common with the frequency positions of features observed in the powder three-pulse ESEEM spectra collected at g ) 2.02.29 Thus, as assigned previously,29 the single narrow feature at ν2H arises from two weakly coupled 2H nuclei. The pair of features that are split symmetrically about ν2H by approximately 1.4 MHz arise from a single strongly coupled 2H nucleus. The intensity in the spectral region 3.5-6.0 MHz arises from two contributions. The positive intensity in the region 3.5-6.0 MHz is the ms ) +1/2 feature from coupling between the electron spin at C2 and one β-2H nucleus bonded to C1 (2Hβ1). Intensity from the ms ) -1/2 conjugate feature is partially obscured by the baseline disturbance at the beginning of the Fourier transform. The depression at frequencies 0.2-0.3 MHz above 2ν2H ) 4.0-4.2 MHz (for B0 ) 306.0-319.0 mT) is the negative phase sum-combination feature of the strongly coupled C5′-methyl 2H nucleus. The shift of the sum combination line from 2ν2H is consistent with the large dipolar contribution to this hyperfine coupling.65 As expected from previous three-pulse

ESEEM studies,29 features from the Hβ2 coupling are too weak to observe above the experimental noise level under the microwave field and magnetic field conditions of 8.8 GHz and 306.0-319.0 mT used in the present experiments. Simulation of the ESEEM. Figure 7 shows simulated quotient two-pulse ESEEM overlaid with the corresponding experimental waveforms. The best-fit parameter set used for the calculated ESEEM was obtained from the global simulation of the ESEEM collected at the four magnetic field values. The best fit parameters and the parameter ranges are listed in Table 2. Simulation fitting parameters (A, k, and associated σt values) are presented in the legend to Figure 7. The ranges in Table 2 were obtained by pooling a list of parameter values and fit numbers for 71 370 fitting attempts (simulation runs), choosing a cutoff fit number, and then listing the range of each parameter that gave a fit number below the cutoff. Plots of fit numbers versus parameter value for the different parameters are included in Supporting Information, Figure S1. Table 2 shows that there are two possible values for the CoII-C5′ distance, 6.3 and 12.5 Å, because, owing to the symmetry of the atom arrangement shown in Figure 3, the CoII-C2 vector can be reversed without altering the ESEEM. The two distances are distinguished by additional considerations, as described below. The Aiso,n and atom-atom distance parameters from the orientation-selection two-pulse ESEEM are comparable with the corresponding values obtained from simulations of the powder three-pulse ESEEM.29 As given in Table 2, the values of Aiso,a ) -0.75 MHz and ra ) 2.42, rb ) 4.07, and rc ) 4.09 Å are the same within error as the values of Aiso,1 ) -0.9 MHz and ra ) 2.3, rb ) rc ) 4.2 Å obtained in the powder three-pulse ESEEM study.29 The Aiso,β1 and rβ1 values in Table 2 are also the same within error as the corresponding values of 4.7 MHz

3062 J. Phys. Chem. B, Vol. 109, No. 7, 2005

Canfield and Warncke

Figure 8. Fourier transforms of the two-pulse 2H/1H quotient ESEEM collected from the product radical in ethanolamine deaminase at four different magnetic field values (black lines), and corresponding overlaid simulations (green lines). Features arising from the strongly coupled 2H (2Hs), the two weakly coupled 2H (2Hw), and the observed β-2H (2Hβ1) are noted. The dark horizontal bars mark the range of frequencies (0.49-9.9 MHz) used in the fitting procedure. The four vertical lines are at the frequencies ν2H ) B0 × 6.53566 MHz/T, ν2H ( 0.7 MHz, and 2ν2H. Experimental conditions: Same as described in the legend of Figure 7. Simulation fitting parameters: All as defined in eq 7 (with A ) 1 and B ) 0) and (A) σf ) 8.95 × 103, (B) σf ) 244, (C) σf ) 304, and (D) σf ) 5.36 × 103.

and 2.5 Å for 2Hβ1 obtained in the powder three-pulse ESEEM study.29 In addition, the sum of the θβ1 and θβ2 values corresponding to the adjustable Aiso,β1 parameter and fixed Aiso,β2 value is 121°, which is consistent with the value of 120° expected for the angular displacement of Hβ1 and Hβ2 about the C1-C2 axis, because C1 is sp3-hybridized.28 A value of 120° was previously obtained in the powder three-pulse ESEEM study.28,29 Figure 8 shows the simulated frequency spectra, obtained by Fourier transforming the ESEEM in Figure 7, overlaid with the corresponding experimental Fourier transforms. For the C5′ 2H couplings, the simulations reproduce the growth and broadening of the feature from the weakly coupled 2H (Hb, Hc, or Hw) centered at ν2H, and the dramatic changes in the asymmetric line shape of the strongly coupled 2H (Ha or Hs), that are associated with the increase in magnetic field from 306.0 to 313.3 mT. The interplay between the magnetic field-dependent β-2H intensity and the overlapping negative phase of the 2Hs sum combination line is also reproduced. Discussion Model for the Geometry of Reactant Centers in the Active Site. Figure 9 shows the model for the geometry of the reactant centers in the active site in the CoII-product radical pair state of ethanolamine deaminase, which is based on the best-fit parameter values presented in Tables 1 and 2. The value of R ) 9.3 Å for the CoII-C2 distance is obtained from the EPR simulations. The global best-fit ESEEM simulation parameters establish the relative positions of CoII, C2, C1, and the C5′methyl group. In the model, C5′ is placed between CoII and C2, which corresponds to selection of the CoII-C5′ distance of 6.3 Å given in Table 2. This is consistent with the shortest-

distance path for the migrating C5′ radical center between CoII and the substrate/product binding site. A position of C5′ between CoII and C2 is also consistent with X-ray crystallographic structures of coenzyme B12-dependent enzymes,10-19 which indicate that the space above the β-face of the cobalamin between cobalt and the substrate is relatively free, whereas displacement of C5′ beyond the bound substrate (or inhibitors or products) is restricted by the protein that forms the side of the substrate binding site distal to CoII. The C5′-methyl and C2-methylene groups are at closest van der Waals contact. C5′ lies slightly off of the CoII-C2 axis, corresponding to a perpendicular distance, r⊥, of 1.6 Å (range 0.1-2.1 Å) and an angle between the CoII-C2 and C5′-C2 axes of 27°. Other notable orientations in the model include the C5′-Ha-C2 angle of 161° (angular range of 143-180°), which indicates that this group of centers is nearly linear. The C1-C2 axis of the product radical makes an angle of 50° (angular range of 46-85°) with the CoII-C2 axis. Implications for the Mechanism of Catalysis. The process of radical pair recombination from the CoII-product radical pair state to the diamagnetic adenosylcob(III)alamin-product state is mediated by migration of the C5′ radical center of the deoxyadenosyl group,29 as inferred earlier from the studies of the substrate radical state.21-23 The model in Figure 9 indicates that C5′ undergoes a displacement during radical pair recombination of at least 4.3 Å, from the position near the product radical to the bonded position 2.0 Å from the cobalt atom66 in adenosylcob(III)alamin. This distance is shorter than the radical pair separation distance of 6.0 Å that was concluded from the orientation-selection ESEEM study of the CoII-substrate radical intermediate.23 The shorter distance for radical migration determined from the CoII-product radical pair state arises from

Product Radical Structure in Ethanolamine Deaminase

J. Phys. Chem. B, Vol. 109, No. 7, 2005 3063 Acknowledgment. Supported by grant R01 DK54514 from the National Institutes of Health. We thank Ms. Lori Anderson for technical assistance, and Professor Dale E. Edmondson (Emory University) for use of fermentation facilities. Calculations were performed, in part, using computer resources at the Emerson Center for Scientific Computation at Emory University, which is funded by National Science Foundation grant CHE0079627 and an IBM Shared University Research Award. Supporting Information Available: Plots of fitting function values for individual parameters. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes

Figure 9. Model for the structure of the reactant centers in the active site of the CoII-product radical pair state in ethanolamine deaminase. (A) View along the line perpendicular to the CoII-C2-C1 plane. (B) View after 90° rotation about the CoII-C2 axis, relative to view in (A). The positions of CoII (dark blue), C1, C2, C5′ (charcoal), and the C5′-methyl hydrogen atoms (light blue) are marked by small spheres. The dotted surfaces represent the extent of the van der Waals radii of the atoms. The positions and orientations of the atom centers are to scale, relative to the plane of the CoII-C2 axis, which lies in the plane of the page. The distance scale corresponds to the CoII-C2 axis on the plane of the page. In (A), Hc, HR1, and Hβ1 are in the foreground, and Hb, HR2 and Hβ2 are in the background. In (B), Hc, HR1, and Hβ1 are near the bottom, and Hb, HR2, and Hβ2 are near the top.

the shorter CoII-C2 distance of 9.3 Å (range of 8.5-10.5 Å), relative to the CoII-C1 distance of 11.1 Å (range of 10.911.5 Å)23 obtained for the CoII-substrate radical pair state. The orientation of the C2-C1 axis relative to the CoII-C2 axis in Figure 9 supports the EPR simulation result that C1 is more distant from CoII than C2. The model presented in Figure 9, along with the assumption that the positions of the C1 and C2 atoms in the substrate and product radical states are comparable, suggests that radical pair separation may occur over a distance that is 1-2 Å farther than radical pair recombination. The small shift of the C5′-methyl group from the position determined in the substrate radical state to the position in the product radical state must occur during or after migration of the amine from C2 to C1. The best-fit [C5′-Ha‚‚‚C2•] angle of 161°, as given in Table 2 and shown in Figure 9, suggests that the second hydrogen atom transfer from the C5′-methyl group to C2 (HT2 in Figure 2) proceeds with a nearly in-line arrangement, and therefore a transition state with predominantly linear character. The approximately perpendicular orientation of the C2-C1 and C5′C2 axes is also consistent with a nearly linear transition state. In this orientation, the p-orbital on C2 that accepts the hydrogen atom is pointing along the C5′-Ha axis. Figure 9 shows that the plane defined by the HR1, HR2, C2, and C1 atoms defines sides of the product radical that face toward or away from CoII. The radical migration and hydrogen atom transfer reactions occur on the side of the plane facing CoII. Amine migration occurs on the side of the plane facing away from CoII. The partitioning of the subreactions into separate spatial regions of the active site would allow specialization of the protein structures that guide these distinctive processes and also eliminate interference between the reactions.

(1) Bandarian, V.; Reed, G. H. Ethanolamine Ammonia-Lyase. In Chemistry and Biochemistry of B12; Banerjee, R., Ed.; John Wiley and Sons: New York, 1999; p 811. (2) Bradbeer, C. J. Biol. Chem. 1965, 240, 4669. (3) B12; Dolphin, D., Ed.; Wiley: New York, 1982; Vol. 2. (4) Buckel, W.; Golding, B. T. Chem. Soc. ReV. 1996, 25, 329. (5) Banerjee, R. Chemistry and Biochemistry of B12; John Wiley and Sons: New York, 1999. (6) Babior, B. M.; Moss, T. H.; Orme-Johnson, W. H.; Beinert, H. J. Biol. Chem. 1974, 249, 4537. (7) Bandarian, V.; Reed, G. H. Biochemistry 2002, 41, 8580. (8) Warncke, K.; Schmidt, J. C.; Ke, S.-C. J. Am. Chem. Soc. 1999, 121, 10522. (9) Licht, S. S.; Lawrence, C. C.; Stubbe, J. J. Am. Chem. Soc. 1999, 121, 7463. (10) Mancia, F.; Keep, N. H.; Nakagawa, A.; Leadlay, P. F.; McSweeney, S.; Rasmussen, B.; Bo¨secke, P.; Diat, O.; Evans, P. R. Structure 1996, 4, 339. (11) Mancia, F.; Evans, P. R. Structure 1998, 6, 711. (12) Mancia, F.; Smith, G. A.; Evans, P. R. Biochemistry 1999, 38, 7999. (13) Reitzer, R.; Gruber, K.; Jogl, G.; Wagner, U. G.; Bothe, H.; Buckel, W.; Kratky, C. Structure 1999, 7, 891. (14) Gruber, K.; Reitzer, R.; Kratky, C. Angew. Chem., Int. Ed. Engl. 2001, 40, 3377. (15) Shibata, N.; Masuda, J.; Tobimatsu, T.; Toraya, T.; Suto, K.; Morimoto, Y.; Yasuoka, N. Structure 1999, 7, 997. (16) Masuda, J.; Shibata, N.; Morimoto, Y.; Toraya, T.; Yasuoka, N. Structure 2000, 8, 775. (17) Yaminishi, M.; Yunoki, M.; Tobimatsu, T.; Sato, H.; Matsui, J.; Dokiya, A.; Iuchi, Y.; Oe, K.; Suto, K.; Shibata, N.; Morimoto, Y.; Yasuoka, N.; Toraya, T. Eur. J. Biochem. 2002, 269, 4484. (18) Liao, D.-I.; Dotson, G.; Turner, I.; Reiss, L.; Emptage, M. J. Inorg. Biochem. 2003, 93, 84. (19) Sintchak, M. D.; Arjara, G.; Kellogg, B. A.; Stubbe, J.; Drennan, C. L. Nature Struct. Biol. 2002, 9, 293. (20) Champloy, F.; Jogl, G.; Reitzer, R.; Buckel, W.; Bothe, H.; Beatrix, B.; Broeker, G.; Michalowicz, A.; Meyer-Klaucke, W.; Kratky, C. J. Am. Chem. Soc. 1999, 121, 11780. (21) Lobrutto, R.; Bandarian, V.; Magnusson, O. T.; Chen, X.; Schramm, V. L.; Reed, G. H. Biochemistry 2001, 40, 9. (22) Warncke, K.; Utada, A. S. J. Am. Chem. Soc. 2001, 123, 8564. (23) Canfield, J. M.; Warncke, K. J. Phys. Chem. B 2002, 106, 8831. (24) Faust, L. P.; Connor, J. A.; Roof, D. M.; Hoch, J. A.; Babior, B. M. J. Biol. Chem. 1990, 265, 12462. (25) Faust, L. P.; Babior, B. M. Arch. Biochem. Biophys. 1992, 294, 50. (26) Ke, S.-C. Biochim. Biophys. Acta 2003, 1620, 267. (27) Boas, J. F.; Hicks, P. R.; Pilbrow, J. R.; Smith, T. D. J. Chem. Soc., Faraday II 1978, 74, 417. (28) Warncke, K.; Canfield, J. M. J. Am. Chem. Soc. 2004, 126, 5930. (29) Warncke, K. Biochemistry, in press. (30) Press, W. H.; Teukolsky, S. A.; Vetterling, W. T.; Flannery, B. P. Numerical Recipes in C: The Art of Scientific Computing, 2nd ed.; Cambridge University Press: New York, 1992. (31) Lagarias, J. C.; Reeds, J. A.; Wright, M. H.; Wright, P. E. SIAM J. Optimization 1998, 9, 112. (32) Rist, G. H.; Hyde, J. S. J. Chem. Phys. 1970, 52, 4633. (33) Hoffman, B. M.; Martinsen, J.; Venters, R. A. J. Magn. Reson. 1984, 59, 110. (34) Hoffman, B. M.; Venters, R. A.; Martinsen, J. J. Magn. Reson. 1985, 62, 537. (35) Hoffman, B. M.; DeRose, V. J.; Doan, P. E.; Gurbiel, R. J.; Houseman, A. L. P.; Telser, J. I., Volume., Eds., pp. Metalloenzyme activesite structure and function through multifrequency cw and pulsed ENDOR.

3064 J. Phys. Chem. B, Vol. 109, No. 7, 2005 In Biological Magnetic Resonance; Berliner, L. J., Reuben, J., Eds.; Plenum Press: New York, 1993; Vol. 13, p 151. (36) Lee, H.-I.; Doan, P. E.; Hoffman, B. M. J. Magn. Reson. 1999, 140, 91. (37) Bandarian, V.; Poyner, R. R.; Reed, G. H. Biochemistry 1999, 38, 12403. (38) Wertz, J. E.; Bolton, J. R. Electron Spin Resonance; Chapman and Hall: New York, 1986. (39) Dikanov, S. A.; Tsvetkov, Y. D. Electron Spin-Echo EnVelope Modulation (ESEEM) Spectroscopy; CRC Press: Boca Raton, FL, 1992. (40) Deligiannakis, Y.; Louloudi, M.; Hadjiliadis, N. Coord. Chem. ReV. 2000, 204, 1. (41) Schweiger, A.; Jeschke, G. Principles of pulse electron paramagnetic resonance; Oxford University Press: New York, 2001. (42) Harkins, T.; Grissom, C. B. J. Am. Chem. Soc. 1995, 117, 566. (43) Kaplan, B. H.; Stadtman, E. R. J. Biol. Chem. 1968, 243, 1787. (44) Hollaway, M. R.; Johnson, A. W.; Lappert, M. F.; Wallis, O. C. Eur. J. Biochem. 1980, 111, 177. (45) Bandarian, V.; Reed, G. H. Biochemistry 1999, 38, 12394. (46) Mims, W. B. J. Magn. Reson. 1984, 59, 291. (47) Eaton, S. S.; More, K. M.; Sawant, B. M.; Boymel, P. M.; Eaton, G. R. J. Magn. Reson. 1983, 52, 435. (48) Canfield, J. M. Approaching magnetic field effects in biology using the radical pair mechanism, University of Illinois, Urbana-Champaign, 1997. (49) Mims, W. B. Phys. ReV. B 1972, 5, 2409. (50) Mims, W. B. Phys. ReV. B 1972, 6, 3543. (51) Ke, S.-C.; Warncke, K. J. Am. Chem. Soc. 1999, 121, 9922. (52) Gordy, W. Theory and Applications of Electron Spin Resonance; John Wiley & Sons: New York, 1980; Vol. XV. (53) Heller, C.; McConnell, H. M. J. Chem. Phys. 1960, 32, 1535. (54) Fessenden, R. W.; Schuler, R. H. J. Chem. Phys. 1963, 39, 2147.

Canfield and Warncke (55) Kevan, L.; Bowman, M. K. Modern Pulsed and Continuous WaVe Electron Spin Resonance; Wiley: New York, 1990. (56) Stoll, S.; Schweiger, A. J. Magn. Reson. 2003, 163, 248. (57) Rowan, L. G.; Hahn, E. L.; Mims, W. B. Phys. ReV. 1965, 137, A61. (58) Kevan, L.; Schwartz, R. N. Time Domain Electron Spin Resonance; John Wiley and Sons: New York, 1979. (59) Mims, W. B.; Davis, J. L.; Peisach, J. J. Magn. Reson. 1990, 86, 273. (60) Pilbrow, J. R. EPR of B12-dependent enzyme reactions and related systems. In B12; Dolphin, D., Ed.; Wiley: New York, 1982; Vol. 1, p 431. (61) Gerfen, G. J. EPR spectroscopy of B12-dependent enzymes". In: Chemistry and Biochemistry of B12. In Chemistry and Biochemistry of B12; Banerjee, R., Ed.; John Wiley and Sons: New York, 1999; p 165. (62) Gerfen, G. J.; Licht, S.; Willems, J.-P.; Hoffman, B. M.; Stubbe, J. J. Am. Chem. Soc. 1996, 118, 8192. (63) Coffman, R. E.; Buettner, G. R. J. Phys. Chem. 1979, 83, 2387. (64) Hyde, J. S.; Rist, G. H.; Eriksson, L. E. G. J. Phys. Chem. 1968, 72, 4269. (65) Astashkin, A. V.; Dikanov, S. A. In AdVanced EPR: Applications in Biology and Chemistry; Hoff, A. J., Ed.; Elsevier: New York, 1989; p 59. (66) Kratky, K.; Krautler, B. X-ray crystallography of B12. In Chemistry and Biochemistry of B12; Banerjee, R., Ed.; John Wiley and Sons: New York, 1999; p 9. (67) Glusker, J. P. X-ray crystallography of B12 and cobaloximes. In B12; Dolphin, D., Ed.; Wiley: New York, 1982; Vol. 1, p 23. (68) Ke, S.-C.; Torrent, M.; Museav, D. G.; Morokuma, K.; Warncke, K. Biochemistry 1999, 38, 12681. (69) Abend, A.; Bandarian, V.; Nitsche, R.; Stupperich, E.; Retey, J.; Reed, G. H. Arch. Biochem. Biophys. 1999, 370, 138. (70) Babior, B. M. Ethanolamine ammonia-lyase. In B12; Dolphin, D., Ed.; Wiley: New York, 1982; Vol. 2, p 263.